Wireless communications are prone to the interference, so the data transmission among the nodes in wireless sensor networks deployed in complex environments has the obvious uncertainty. This paper adopts probability theory to extend the existed interference model, and gives an interference analysis model and implements it through the cross-layer method. In addition, the isotonic property of the interference-aware routing metric is proved. Then, a probabilistic routing algorithm is proposed and its correctness and time-space complexity are analyzed. Simulation results show that the proposed algorithm can achieve better packet delivery ratio, throughput, jitter and average delay in dense deployment under the different loads at the expense of the comparable average length of paths compared with the Adhoc On-Demand Distance Vector (AODV) algorithm.